Skip to content

docs: warn about structured generation without a prompt for Llama tokenizers #1929

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
pctablet505 wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

docs: warn about structured generation without a prompt for Llama tokenizers #1929

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions docs/features/core/generator.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,6 +81,15 @@ book = BookRecommendation.model_validate_json(result)
print(f"{book.title} by {book.author} ({book.year})")
```

!!! warning "Structured generation without a prompt"

Llama/SentencePiece-based tokenizers treat leading spaces specially. When
structured generation starts from an empty prompt, Outlines decodes tokens
strictly, so a token that normally represents a leading-space word (e.g.
`" https"`) can be rejected even though the same token is valid after a
prompt. Provide a prompt when using structured generation with these
tokenizers to avoid spuriously invalid outputs.

## Parameters

- `model`: The language model to use for generation
Expand Down